An infrared spectroscopic study of the nature of zinc carboxylates in oil paintings

Abstract

The formation of metal soaps is a major problem for oil paintings conservators. The complexes of either lead or zinc and fatty acids are the product of reactions between common pigments and the oil binder, and they are associated with many types of degradation that affect the appearance and stability of oil paint layers. Fourier Transform Infrared spectroscopy (FTIR) reveals that a paint sample from The Woodcutter (after Millet) by Vincent van Gogh contains two distinct zinc carboxylate species, one similar to zinc palmitate and one that is characterized by a broadened asymmetric stretch COO band shifted to 1570-1590 cm−1 . Though this observation has been made in many paintings, it was previously left unexplained. It is shown that neither variations in the composition of zinc soaps (i.e. zinc soaps containing mixtures of fatty acids or metals) nor fatty acids adsorbed on pigment surfaces are responsible for the second zinc carboxylate species. X-ray diffraction (XRD) and FTIR analysis indicate that the broad COO band represents amorphous zinc carboxylates. These species can be interpreted as either non-crystalline zinc soaps or zinc ions bound to carboxylate moieties on the polymerized oil network, a system similar to ionomers. These findings uncover an intermediate stage of metal soap-related degradation of oil paintings, and lead the way to improved methods for the prevention and treatment of oil paint degradation.

Full text

An infrared spectroscopic study of the nature of
zinc carboxylates in oil paintings†
Joen J. Hermans,*Katrien Keune, Annelies van Loon and Piet D. Iedema
The formation of metal soaps is a major problem for oil paintings conservators. The complexes of either lead
or zinc and fatty acids are the product of reactions between common pigments and the oil binder, and they
are associated with many types of degradation that affect the appearance and stability of oil paint layers.
Fourier transform infrared spectroscopy (FTIR) reveals that a paint sample from The Woodcutter (after
Millet) by Vincent van Gogh contains two distinct zinc carboxylate species, one similar to crystalline zinc
palmitate and one that is characterized by a broadened asymmetric stretch COO

band shifted to 1570–
1590 cm
1
. This observation has been made in many paintings. Although several hypotheses exist to
explain the shifted broad carboxylate band, these were not supported by experimental evidence. In this
paper, experiments were carried out to characterize the second zinc carboxylate type. It is shown that
neither variations in the composition of zinc soaps (i.e. zinc soaps containing mixtures of fatty acids or
metals) nor fatty acids adsorbed on pigment surfaces are responsible for the second zinc carboxylate
species. X-Ray diffraction (XRD) and FTIR analysis indicate that the broad COO

band represents
amorphous zinc carboxylates. These species can be interpreted as either non-crystalline zinc soaps or
zinc ions bound to carboxylate moieties on the polymerized oil network, a system similar to ionomers.
These findings uncover an intermediate stage of metal soap-related degradation of oil paintings, and
lead the way to improved methods for the prevention and treatment of oil paint degradation.
1 Introduction
From a chemical point of view, oil paintings are not stable
objects. The drying oil that acts as binding medium in the oil
paint, usually linseed oil, polymerizes in a matter of weeks,
1
but
long-term degradation processes take place over the course of
centuries that affect the appearance and structural integrity of
the painting. A prominent issue for paintings conservators is
the formation of metal soaps. The presence of metal soaps has
been reported for numerous paintings ranging from Rembrandt
in the 17th century to Salvador Dal´
ıin the 20th century.
2–11
These complexes of either lead or zinc with stearic and/or pal-
mitic acid are the consequence of reactions between the
common pigments lead white (2PbCO
3
$Pb(OH)
2
), red lead
(Pb
3
O
4
), lead–tin yellow (Pb
2
SnO
4
) or zinc white (ZnO) and the
oil binder. Metal soaps defects may appear in the paint system
in many different forms: as large aggregates that deform paint
layers, as deposits on the surface of a paint or homogeneously
spread throughout paint layers. Besides causing a loss of
pigment and a change in surface texture, the formation of metal
soaps has been linked to cases of brittleness, loss of strength
and delamination of paint layers.
The variation in metal soap appearance and location within
the paint suggests a complex set of processes by which metal
soaps form and separate from the paint matrix. However, the
chemical reactions that lead to metal soap formation and the
transport mechanisms for the reactants and products in these
reactions are not fully understood. Answering these questions is
a challenging task because of the limited availability of sample
material from real paintings and the difficulty of reproducing
all the different states of degradation that a painting might be
in. A full understanding of metal soap related phenomena in oil
paintings must start with an accurate analysis of the molecular
structure of metal soaps and all the variations that might be
found in oil paint systems. Ultimately, we need to be able to
describe the composition and structure of an oil paint on a
molecular level in terms of, for instance, concentrations of
functional groups, polymerization degree, cross-link density
and metal content. Only then, it is possible to investigate how
paint composition and environmental factors affect the degree
of metal soap related degradation of a painting and nally to
develop conservation strategies that minimize the chance of
further paint degradation.
Fourier transform infrared (FTIR) spectroscopy is a powerful
technique to identify metal soaps in cross-sections of oil
Van 't HoffInstitute for Molecular Sciences, University of Amsterdam, PO box 94157,
1090GD Amsterdam, The Netherlands. E-mail: j.j.hermans@uva.nl; Tel: +31 (0)20 525
6442
†Electronic supplementary information (ESI) available: Time-dependent
ATR-FTIR spectra and XRD traces of the ZnUFA complex, d-spacing of ZnC
16
C
18
mixtures, and additional FTIR spectra of Zn-pol and ZnO-LO systems. See DOI:
10.1039/c5ja00120j
Cite this: J. Anal. At. Spectrom.,2015,
30,1600
Received 2nd April 2015
Accepted 30th April 2015
DOI: 10.1039/c5ja00120j
www.rsc.org/jaas
1600 |J. Anal. At. Spectrom.,2015,30,1600–1608 This journal is © The Royal Society of Chemistry 2015
JAAS
PAPER
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online
View Journal
| View Issue

paintings. The position and shape of the asymmetric stretch
vibration band of the carboxylate group in the metal soap (n
a
-
COO

) is characteristic for the type of metal ion, e.g. a single
sharp band at 1536 cm
1
for zinc soaps.
12,13
Though it is known
that metal soaps formed in oil paint layers contain mostly
stearic and palmitic acid,
4,12
the FTIR spectra of metal soaps
formed in paints oen do not match with the metal stearates or
palmitates synthesized as references. Specically, the n
a
COO

band is frequently signicantly broadened and shied to higher
wavenumbers (between 1570 and 1590 cm
1
) in ZnO containing
paint layers.
7,9,14–17
We discuss Attenuated Total Reection
Fourier Transform infrared microscopy (m-ATR-FTIR) analysis of
a sample from the painting De Houthakker (naar Millet) (‘The
Woodcutter (aer Millet)’) by Vincent van Gogh as a clear
example of this phenomenon.
Several explanations have been suggested to account for the
variation in the n
a
COO

band in zinc white paints. In crystalline
zinc palmitate (Zn(C
16
)
2
), the zinc atoms are surrounded by four
equivalent carboxylate groups,
18
giving rise to a single sharp
n
a
COO

band in the FTIR spectrum. The broadening of the
carboxylate band suggests that there is increased variation in
the local environment of the carboxylate groups. We have
synthesized a range of zinc soaps that contain various mixtures
of carboxylic acids or mixtures of metals to study the effect of
added complexity in the zinc soap structure on the carboxylate
coordination. Secondly, the adsorption of fatty acids on ZnO
surfaces has been suggested as an explanation for the broad
n
a
COO

band in FTIR spectra.
14,15
The effect of different binding
modes of carboxylic acids on the surface of ZnO particles has
been studied by Lenz et al.
19
While it is likely that the energies of
COO

vibrations are not identical in Zn(C
16
)
2
and fatty acids
adsorbed on ZnO surfaces, it is not clear whether the concen-
tration of surface-bound fatty acids is large enough to cause the
entire n
a
COO

band to shiin paint layers that have a high
concentration of zinc carboxylates. We investigated this effect
by following the reaction of ZnO powder in a solution of
palmitate ions. Lastly, we considered the possibility that there is
disorder in the alkyl chains of zinc soap phases that form in
ZnO containing paints, by studying the crystallinity of zinc
carboxylates formed in a model paint system and zinc con-
taining linseed oil polymers. Dreveni et al. found that the
position and shape of the COO

vibration bands are strongly
dependent on the ability of the alkyl chains to interact and pack
into a well-ordered lattice.
20
While the fatty acid chains are
neatly packed in an all-trans fashion in crystalline Zn(C
16
)
2
,
18,21
chains might not have sufficient mobility to pack in a well-
ordered manner when zinc soaps form in a polymerized oil
network.
This work aims to give a complete overview of the different
molecular structures of zinc carboxylates that may be found in
oil paint samples. Applying a combination of Fourier transform
infrared spectroscopy (ATR-FTIR) and X-ray diffraction (XRD),
the effect of relatively minor changes to metal soap composition
or physical state can be investigated and detailed structures can
be resolved. Finally, we applied the obtained information on the
likely structures of zinc carboxylates to develop an hypothesis
on the different stages of metal soap formation in oil paint and
the transport processes that are involved.
2 Experimental
2.1 Synthesis of zinc soaps
Zinc palmitate (Zn(C
16
)
2
) and zinc palmitate/stearate (ZnC
16
C
18
)
were synthesized by mixing basic aqueous solutions of the fatty
acids with a solution of zinc nitrate. Mixed-metal soaps were
prepared by using a solution of NaOH to dissolve palmitic acid
(ZnNa
2
(C
16
)
4
), or by melting stochiometric amounts of Zn(C
16
)
2
and KC
16
under inert atmosphere (ZnK
2
(C
16
)
4
). All these
procedures are described in detail in ref. 18. All zinc soaps were
thoroughly dried before analysis either by placing samples in an
oven at 110 C or in a desiccator over P
2
O
5
, and their purity was
conrmed by FTIR and XRD. In experiments where the reaction
between ZnO and palmitic acid was followed, ZnO powder was
added to an aqueous solution of palmitic acid containing an
excess of triethylamine. Samples were taken aer 15 seconds
and then every 2 minutes. The solid fractions were immediately
separated by vacuum ltration and dried in an oven at 110 C.
To synthesize zinc soaps of unsaturated fatty acids (UFAs),
cold-pressed untreated linseed oil (Kremer Pigmente) was
hydrolyzed using an excess of a concentrated aqueous solution
of KOH. Aer neutralization with concentrated HCl, the fatty
acid mixture was extracted with dichloromethane and dried
with MgSO
4
. Evaporation of the solvent yielded a clear brown
liquid mixture of 53% linolenic, 19% linoleic, 16% oleic, and
12% stearic and palmitic acid (composition determined by
1
H-
NMR and
13
C-NMR). Zinc soaps of this fatty acid mixture
(ZnUFA) were prepared by mixing stoichiometric amounts of
zinc nitrate and UFAs in a 1 : 1 mixture of ethanol and diethyl
ether containing and excess amount of triethylamine. Aer
approximately one hour, the solvent was evaporated to yield an
off-white product with a pasty texture.
2.2 Preparation of model paint lm (ZnO-LO)
The composition of the model paint system was adapted in an
attempt to promote the rapid formation of zinc carboxylates, i.e.
a high oil/pigment ratio and an addition of water to the paint
mixture. Model paint lms were prepared by stirring 3.0 g cold-
pressed untreated linseed oil (Kremer Pigmente) with 0.5 g ZnO
(Sigma-Aldrich) and 1 mL demineralized water in a sealed vial at
room temperature for three days. Aer allowing the mixture to
settle, a few drops of the oil layer were spread on a glass slide
and leto dry in air at room temperature for up to seven weeks.
2.3 Preparation of zinc ionomer (Zn-pol)
Zinc sorbate was prepared by dissolving 550 mg sorbic acid with
1 mL triethylamine in 20 mL demineralized water at 50 C. The
addition of 1.0 g Zn(NO
3
)
2
$6H
2
O dissolved in 5 mL water
resulted in immediate precipitation of the white product. Aer
stirring for 20 minutes, the product was separated by vacuum
ltration and dried over P
2
O
5
.
A ionomer lm was made by mixing 55 mg zinc sorbate with
200 mg cold-pressed untreated linseed oil, and spreading the
This journal is © The Royal Society of Chemistry 2015 J. Anal. At. Spectrom.,2015,30,1600–1608 | 1601
Paper JAAS
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

resulting turbid paste thinly on a glass plate. Aer curing at
150 C in air overnight, a homogeneous yellow transparent
polymer lm was obtained.
2.4 Analysis
A small sample of The Woodcutter (aer Millet) by Vincent van
Gogh (Fig. 1) was embedded in a polyester resin block. The resin
block was sanded down and polished until the cross-section was
situated on the surface of the block.
Optical analysis was carried out using a Zeiss Axioplan 2
equipped with both a polarized light source and a UV light
source (365 nm).
Scanning electron microscopy was performed with a JEOL
JSM 5910 LV microscope. Backscattered-electron images (SEM-
BSE) were mostly taken at 20 kV accelerating voltage at a 10 mm
eucentric working distance. Samples were gold coated to
improve surface conductivity.
Cross-sections were analyzed with m-ATR-FTIR using a Perkin
Elmer Spectrum 100 FTIR spectrometer combined with a
Spectrum Spotlight 400 FTIR microscope equipped with a 16 
1 pixel linear MCT array detector at 8 cm
1
resolution. A Perkin
Elmer ATR Ge crystal accessory was used for ATR imaging.
Spectra were collected in a 600–4000 cm
1
range and averaged
over 2 scans.
ATR-FTIR spectra of bulk material were collected on a Varian
660-IR FT-IR spectrometer combined with a Pike Technologies
diamond GladiATR unit with 4 cm
1
resolution. Spectra were
collected in a 600–3500 cm
1
range and averaged over 16 scans.
XRD was performed on a Rigaku MiniFlex II desktop X-ray
diffractometer with Cu Karadiation (l¼1.54180 ˚
A) at 30 kV and
15 mA. The equipment was tted with a Ni Kbsuppression
lter. Diffractograms were recorded in a 2q¼1–40range
(2.5min
1
scan rate and 0.025step size). Powder samples
were nely ground with mortar and pestle, and manually
pressed into glass sample holders. Model paint lms prepared
on glass slides were lied from their support and measured on
the underside, since a transparent oil layer formed on top of the
lm.
3 Results and discussion
3.1 Painting cross-section analysis
The painting The Woodcutter (aer Millet) was painted by Vin-
cent van Gogh in 1889 (Fig. 1). A small sample was taken of the
light green paint near the top of the painting. Images of this
sample obtained with optical microscopy and SEM microscopy
are shown in Fig. 2. A previous study showed that the paints van
Gogh used in this section of the painting contain a large variety
of pigments. In the ground layers, a mixture of lead white,
calcium carbonate, barium sulfate, ochre pigments and a little
carbon black was found. The thick light green paint layer shown
here contains mostly ZnO, and a mixture of emerald green
(Cu(CH
3
COO)
2
$3Cu(AsO
2
)
2
) and chrome green (Cr
2
O
3
)or
viridian (Cr
2
O
3
$2H
2
O).
22
Analysis of the paint cross-section with m-ATR-FTIR revealed
that the light green paint layer contains a high concentration of
zinc carboxylates, as indicated by n
a
COO

bands in the 1500–
1600 cm
1
region. There was considerable variation in the
position and width of the carboxylate bands. Fig. 2c shows a
map of integrated intensity of the sharp band at 1536 cm
1
.A
representative spectrum made by averaging the spectra in area 2
is shown in Fig. 3. This spectrum closely resembles a reference
Fig. 1 De Houthakker (naar Millet) (‘The Woodcutter (after Millet)’)by
Vincent van Gogh, 1889, 44 cm 26.2 cm, van Gogh Museum (Vin-
cent van Gogh Foundation). X marks the spot where the sample shown
in Fig. 2 is taken.
Fig. 2 Paint cross-section from The Woodcutter (after Millet) by
Vincent van Gogh, showing a thick light green pain rich in ZnO on top
of two ground layers. (A) Optical microscopy image, (B) SEM-BSE
image and (C) m-ATR-FTIR map of the sharp band at 1536 cm
1
(integrated between 1495–1546 cm
1
, red color signifies high inten-
sity). Numbers 1 and 2 mark the averaged areas of the FTIR spectra in
Fig. 3.
1602 |J. Anal. At. Spectrom.,2015,30,1600–1608 This journal is © The Royal Society of Chemistry 2015
JAAS Paper
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

spectrum of zinc palmitate, with a sharp n
a
COO

band at 1536
cm
1
and well-dened dCH
2
and n
s
COO

bands at 1450 cm
1
and 1400 cm
1
, respectively. Zinc carboxylate material with a
sharp n
a
COO

band at 1536 cm
1
seems to be more prominent
in the lower parts of the paint layer, and almost absent in top
20–30 mm of the paint layer. Throughout the paint layer
however, a broad n
a
COO

band was found with a maximum at
1570 cm
1
, area 1 shown in Fig. 3. These spectra suggest that
the entire ZnO paint layer is lled with a zinc carboxylate
species that causes a broad n
a
COO

band, while deposits of zinc
soap material comparable to zinc palmitate or stearate refer-
ences are concentrated in the lower section of the paint lm.
In the next sections, we explore possible variations in zinc
carboxylate structures that could induce a change in carboxylate
coordination and cause a broadened and shied n
a
COO

band.
3.2 Zinc soaps of varying composition
Pure zinc palmitate and zinc stearate are most oen used as
reference compounds for FTIR analysis. It is known, however,
that both fatty acids occur simultaneously in metal soap
aggregates in paint samples.
4
Therefore, a series of zinc soaps
containing a mixture of palmitate and stearate in varying ratios
was prepared (ZnC
16
C
18
). A single set of long spacing peaks in
XRD analysis indicated that zinc palmitate and zinc stearate
form a solid solution at all palmitate/stearate ratios (ESI:
Fig. S1†). The FTIR spectra of these mixed fatty acid soaps were
highly similar for all the palmitate/stearate ratios; a typical
spectrum is shown in Fig. 4a. The mixing of palmitate and
stearate only affected the number and intensity of alkyl chain
progression bands in the 1100–1310 cm
1
region.
13
The asym-
metric carboxylic stretch vibration at 1536 cm
1
was unaffected
by this mixing.
We have also explored the effect of unsaturations in the fatty
acid chains on the structure of zinc soaps. Linseed oil typically
contains only 9–13% saturated fatty acids.
1
In paint samples
however, oen only zinc soaps of saturated fatty acids are
found. It is thought that this occurs because in a fully poly-
merized oil lm, hydrolysis of ester bonds in the triglycerides
leaves only the saturated fatty acids ‘free’to diffuse through the
oil network. We have fully hydrolyzed linseed oil to yield a
mixture of mostly unsaturated fatty acids. The zinc complex of
this fatty acid mixture is a crystalline material, as indicated by
the presence of progression bands in the FTIR spectrum in the
1100–1310 cm
1
region (Fig. 4b) and a strong series of long
spacing peaks in the XRD trace (ESI: Fig. S2†). This mono-
crystallinity is quite surprising, given the low melting point of
unsaturated fatty acids and the heterogeneity of the fatty acid
mixture. With a split n
a
COO

band at 1545 cm
1
and 1524
cm
1
, the FTIR spectrum is very similar to that of zinc oleate
and zinc linoleate.
12
Possibly, the presence of cis double bonds
in the unsaturated fatty acid chains causes a slight asymmetry
in the tetrahedral coordination of carboxylic oxygens around
the zinc atoms. However, this minor effect is not capable of
explaining the extensive band broadening that was observed in
the van Gogh paint sample.
Additionally, we investigated whether the structure of ZnUFA
soaps changes as the non-conjugated double bonds in the fatty
acid chains undergo auto-oxidation reactions with atmospheric
oxygen to form peroxides, hydroxides and cross-links.
23
A 0.5
mm layer of ZnUFA was leto cure in air at room temperature
for up to seven months. In that time, the nC]CH band at 3008
cm
1
decreased and a broad –OH band appeared around 3400
cm
1
. The n
a
COO

band was not affected by ageing (ESI:
Fig. S3†). XRD measurements showed clearer changes to the
structure, with a decreasing order in the carbon chain packing
and a small but signicant decrease in the long spacing of 0.6 ˚
A
(ESI: Fig. S2 and S4†).
The last variation in the composition of zinc soaps we
investigated is the incorporation of multiple metals. It was
found that a different type of zinc soaps can form in linseed oil
when a source of sodium or potassium is present such as NaOH
or KCl. The synthesis and structure of these alternative soap
complexes (ZnNa
2
(C
16
)
4
and ZnK
2
(C
16
)
4
) is described in detail in
Fig. 3 Averaged m-ATR-FTIR spectra measured at the two areas in the
cross-section of a zinc white paint layer of The Woodcutter (after
Millet) by Vincent van Gogh, as shown in Fig. 2. The grey area marks the
wavenumber range integrated to create the zinc carboxylate map in
Fig. 2. The bottom spectrum of zinc palmitate is included as a
reference.
Fig. 4 ATR-FTIR spectra of (a) ZnC
16
C
18
(1 : 1 ratio), (b) ZnUFA and (c)
ZnNa
2
(C
16
)
4
.
This journal is © The Royal Society of Chemistry 2015 J. Anal. At. Spectrom.,2015,30,1600–1608 | 1603
Paper JAAS
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

ref. 18. Concerning the FTIR spectrum of these mixed-metal
soaps, there is a major shiof the n
a
COO

band from 1540
cm
1
to 1595 cm
1
with two bands of lower intensity appearing
at 1609 cm
1
and 1620 cm
1
for both the sodium and the
potassium containing complexes (Fig. 4c). Though the band
maximum is shied it is still sharp, in contrast to the band
broadening observed in the paint sample of van Gogh. In
comparison to Zn(C
16
)
2
, this band shiis caused by a transition
from bidentate to effectively monodentate carboxylic groups.
24
The split in the n
a
COO

bands is an effect of the inequivalence
of the carboxylic headgroups in the mixed-metal structure, in
which half the carboxylate groups bind a zinc and a sodium or
potassium atom and half the groups bind two sodium or
potassium atoms. Though the mixed-metal soaps can easily be
synthesized in a reaction mixture of ZnO, palmitic acid and
NaOH/KOH in linseed oil, they have not yet been identied in
oil painting samples.
Concluding, none of the variations in zinc soap composition
we investigated give a satisfying explanation of the n
a
COO

band shiand broadening that is oen observed in zinc white
paint layers.
3.3 ZnO particles in fatty acids solutions
We performed experiments where ZnO powder was added to an
aqueous palmitate solution, in an attempt to measure IR
absorbance of surface-adsorbed carboxylate molecules.
Samples were taken at regular time intervals from 15 s aer the
start of the reaction and analyzed with ATR-FTIR spectroscopy.
Fig. 5 shows that the characteristic zinc palmitate bands are
indeed rising as the reaction progresses. However, the position
of the n
a
COO

band is not affected by the extent of zinc soap
formation. Even aer just 15 seconds reaction time, when the
amount of zinc soap formed is very small and the strength of
absorption is very weak, the band has an absorption maximum
at 1536 cm
1
. This observation has three possible explanations:
the n
a
COO

band of fatty acids adsorbed on ZnO surfaces
lies at 1536 cm
1
;
the fatty acid layer directly on the ZnO surface has shied
n
a
COO

bands, but as soon as IR absorption becomes
measurable, the zinc soap phases are already several layers
thick and the signal is dominated by the non-shied bulk zinc
soap signal;
zinc soap formation does not take place on the ZnO
surface, but instead it is a precipitation reaction between dis-
solved Zn
2+
ions and fatty acids in solution.
We did not investigate which of these explanations is valid,
though Taheri et al. did nd that myristic acid molecules
adsorbed on oxidized zinc surfaces consistently showed sharp
n
a
COO

bands at 1536 cm
1
.
25
However, the situation might be
different for an oil paint system where there is more variation in
the molecules with carboxylate functionalities. While we cannot
conclude that IR absorption of carboxylates directly adsorbed
on a zinc oxide surface is identical to bulk zinc soaps, this
experiment does show that if such surface adsorbed species
exist in paint systems, their concentration is likely to be too
small to account for the entire intensity of the strong n
a
COO

band that is oen measured. In other words, most carboxylates
will be residing in bulk zinc soap material, where the effect of
the pigment surface does not play a role.
Moreover, we observed the broad shied n
a
COO

band in
zinc-containing polymerized oil lms in which pigment was
entirely absent. These crucial experiments are discussed below.
3.4 Disorder in zinc soaps
Next, we discuss disorder in zinc soap structures as an expla-
nation of variations in carboxylate coordination. As mentioned
before, in a polymerized oil network fatty acid chains that are to
become part of the zinc soap phases might not have sufficient
mobility to pack in a well-ordered manner. To investigate
whether disorder in the alkyl chain packing affects the coordi-
nation of carboxylate groups around the zinc ion, we rst
considered the FTIR spectrum of molten zinc palmitate
(Fig. 6b). In the melt, the alkyl chains are completely disordered,
but each carboxylate group is still bound to zinc. Upon melting,
the single n
a
COO

band observed in crystalline zinc palmitate at
1536 cm
1
splits into three bands at 1545, 1593 and 1633 cm
1
at 160 C, when the complex appears as a clear colorless liquid.
Ishioka and co-workers have studied the structural changes
associated with this phase transition with EXAFS and concluded
that the coordination number and Zn–O bond lengths are the
same in both phases.
26
Therefore, the striking changes in the
FTIR spectrum of molten zinc palmitate must be due to a
geometrical distortion of the tetrahedral coordination structure
of the carboxylate groups around the zinc ion, caused by the
increased mobility of the molten alkyl chains. The effect of
complete disorder in the alkyl chains on the n
a
COO

vibration is
signicant. Though the shape of the bands is still different, the
chain disorder produces a shito higher wavenumbers
comparable to the spectrum in area 1 from the van Gogh paint
sample (compare Fig. 6b and c).
In an attempt to produce disordered zinc soaps in a way
more closely related to actual oil paint, we prepared a ZnO
‘paint’lm designed to promote the fast formation of zinc
Fig. 5 ATR-FTIR spectra of ZnO immersed in an aqueous palmitate
solution for 15 seconds to 10 minutes.
1604 |J. Anal. At. Spectrom.,2015,30,1600–1608 This journal is © The Royal Society of Chemistry 2015
JAAS Paper
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

soaps by using an excess of linseed oil and stirring the paint
with water before application. Already aer one week, a broad
n
a
COO

band was observed. Fig. 6a shows the ATR-FTIR spec-
trum of the bottom of such a lm aer 80 days of drying on a
glass support (ZnO-LO). The broad n
a
COO

band centered at
1575 cm
1
is similar to the band from the van Gogh paint in
Fig. 3 (area 1), though a weak shoulder can be observed around
1540 cm
1
. The intensity of the n
a
COO

band relative to the
ester carbonyl band of the cured oil network at 1739 cm
1
indicates that the concentration of zinc carboxylates in this
system is quite high.
It should be remarked that the similarity between the broad
n
a
COO

bands in the van Gogh paint and in ZnO-LO is a crucial
nding (Fig. 6a and c). It implies that this experiment is one of
the few successful attempts to produce a lm containing zinc
carboxylates with a coordination structure very similar to those
found in old paintings on a very short timescale.
15
We pro-
ceeded with X-ray diffraction analysis to study the crystalline
order in this system.
Fig. 7 shows the XRD trace measured on the bottom of the
model ZnO paint lm ZnO-LO, compared to a linseed oil lm
cured with no additives (LO) and a mixture of zinc palmitate
and ZnO. The three characteristic peaks of at 31.8, 34.5and
36.3in ZnO-LO show that there is still ZnO present. The broad
peak with a maximum around 20(d¼4.4 ˚
A) appears in both
the pure linseed oil lm and in the model paint. Such a broad
peak is oen seen in amorphous polymer material and can be
associated with the average distance between the carbon chains
in the cross-linked oil network.
27
In Fig. 7b at small angles we see the typical series of sharp
peaks associated with the long spacing in bulk crystalline zinc
palmitate. In the trace of ZnO-LO, similar peaks appear in the
same region at 1.5, 3.3and around 5, though they are
signicantly broader and much weaker (see inset in Fig. 7).
These three evenly spaced peaks correspond to the rst three
diffraction orders of the long spacing in zinc soaps. The width
of the peaks shows that the crystalline domains are much
smaller than in bulk zinc palmitate. The absence of a well-
resolved series of long spacing peaks prevents an accurate
calculation of the long spacing, but based on this data the value
can be estimated to be around 50 ˚
A. This spacing is signicantly
larger than the long spacing of a zinc soap that contains both
palmitate and stearate (ZnC
16
C
18
, 41.4 ˚
A for a 1 : 1 mixture) or
zinc soaps containing a mixture of unsaturated fatty acids
(ZnUFA, 41.8 ˚
A).
This XRD analysis shows that the ZnO-LO sample contains a
very low concentration of small (semi)-crystalline zinc soap
particles. This small amount cannot, however, account for the
strong n
a
COO

band in the FTIR spectrum of ZnO-LO (Fig. 6).
The lack of strong long spacing peaks leads to the conclusion
that the zinc carboxylate species associated with the broad n
a
-
COO

band around 1570 cm
1
must be amorphous.
Based on the FTIR spectrum and XRD analysis of ZnO-LO, we
will now discuss explanations of the observed n
a
COO

band
broadening and shi. Possibly, the zinc carboxylate species are
zinc soaps of saturated fatty acids being delayed or hindered in
their crystallization process. In a polymerized oil network,
thermal movement of fatty acid chains might be constrained by
the surrounding oil network, making alignment and proper
crystallization of the chains a much slower process than the
coordination of zinc ions by fatty acid carboxylate groups. In
this scenario, the amorphous state of the fatty acid chains
distorts the ideal tetrahedral coordination of the carboxylic
groups around the zinc atom in a fashion similar to molten zinc
palmitate (Fig. 6). If this hypothesis is correct, on a longer
timescale it is expected that the degree of crystallization will
increase and that a sharp n
a
COO

band at 1536 cm
1
will
appear in the FTIR spectrum of ZnO paint layers. This
Fig. 6 ATR-FTIR spectra of (a) zinc soaps formed in a cured film of
ZnO in linseed oil (ZnO-LO), (b) molten zinc palmitate at 160 C, (c)
area 1 in the sample from van Gogh's The Woodcutter (after Millet) and
(d) zinc palmitate at room temperature.
Fig. 7 XRD traces of (a) ZnO-LO, (b) crystalline Zn(C
16
)
2
synthesized
from ZnO and (c) a pure linseed oil film. Intensities of the three traces
are not to scale. The insert shows the three peaks (marked with C)in
the low angle section of the XRD trace measured at higher resolution
and slower scanning speed.
This journal is © The Royal Society of Chemistry 2015 J. Anal. At. Spectrom.,2015,30,1600–1608 | 1605
Paper JAAS
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

explanation implies that the observed system is not in ther-
modynamic equilibrium and is bound to move to a nal state
of, probably, large well-ordered metal soap crystals.
A second interpretation is that the zinc carboxylate species is
not an isolated molecular species present in the paint lm, but
in fact integral part of the oil network structure. Network
carboxylic acid groups may form by hydrolysis of triacyl glycerol
ester bonds aer polymerization of the fatty acid chains has
occurred, or through b-scission reactions and subsequent
oxidation in the unsaturated fatty acid chains.
23
The view of the
oil network as a polymer chain with carboxylic acid side-chains
would make it a complicated type of ionomer. The term “ion-
omer”refers to ion-containing polymers, typically consisting of
a hydrocarbon backbone with a small number of pendant acid
groups. In commercial ionomers, these acid group are oen
partially or completely neutralized by Na
+
,Zn
2+
or other metal
ions. The structure of ionomers has been extensively studied. It
was found that the ionic groups have a well-dened local
structure and cluster to form ion-rich domains within the
polymer matrix.
28–30
Interestingly, an ethylene–methacrylic acid
copolymer completely neutralized by Zn
2+
ions shows a rela-
tively broad n
a
COO

band at 1585 cm
1
, which is similar to the
band we observed for the ZnO-LO system.
28
In order to test the ionomer interpretation, we attempted to
produce a zinc ionomer from linseed oil. Commonly, ionomers
are prepared by letting a polymer that contains ionic groups
react with a metal salt in the melt or in solution. Since poly-
merized linseed oil decomposes before it melts and has a poor
solubility in organic solvents, we chose to introduce zinc ions in
the polymer matrix by copolymerization. The zinc complex of
sorbic acid (2,4-hexadienoic acid, or C6:2) was mixed with
linseed oil and cured as a lm at 150 C in air to induce cross-
linking between the double bonds of unsaturated triglycerides
and the sorbate chains. While zinc sorbate did not dissolve in
linseed oil, upon curing the mixture of zinc sorbate and linseed
oil formed a completely transparent polymer lm (Zn-pol), as
shown in Fig. 8b. Fig. 8a compares the FTIR spectra of zinc
sorbate and Zn-pol. Upon curing, the three sharp bands of zinc
sorbate at 1515, 1620 and 1645 cm
1
disappear and a broad
n
a
COO

band appears with a maximum at 1585 cm
1
. The
absence of a C]CH band at 3008 cm
1
indicated that all double
bonds in the system were either cross-linked or oxidized (ESI:
Fig. S5†). The XRD trace in Fig. 8c conrms that Zn-pol is
amorphous and contains no remaining traces of unreacted zinc
sorbate, showing only the broad peak around 20that was also
found in pure linseed oil lms. The weak maximum around 6
(dz15 ˚
A) proved poorly reproducible and remains to be
interpreted.
The FTIR spectrum of the Zn-pol ionomer is strikingly
similar to the spectra of both the ZnO-LO system and the van
Gogh paint sample, showing that an amorphous polymer
system containing network-bound zinc carboxylate groups is a
valid explanation for the shied broad n
a
COO

band in oil
paints. Moreover, the complete absence of ZnO pigment in this
system emphasizes that the zinc carboxylate species corre-
sponding to the broad n
a
COO

band is not associated with
pigment surfaces. This conclusion is further supported by the
observation that the broad n
a
COO

band in the ZnO-LO system
also appeared in transparent sections of the paint lm that
contained no ZnO pigment (ESI: Fig. S6†).
Applying our ndings to the sample from The Woodcutter
(aer Millet) shown in Fig. 2c, it means that amorphous zinc
carboxylate species are homogeneously spread throughout the
light green paint layer. Crystallization of zinc soaps, as indi-
cated by the sharp 1536 cm
1
band, has occurred mostly in the
lower section of the paint layer, and is concentrated in domains
rich in zinc soaps. A similar distribution of zinc carboxylate
species was also found by Osmond et al. in 40 year old ZnO tube
paint lms and home-made reconstructions,
14
and could be due
to an uneven degree of polymerization in the paint layer or an
inhomogeneous concentration of ‘free’fatty acids.
Based on the results shown here, a clear distinction between
disordered zinc soaps or a zinc-neutralized ionomer cannot be
made. However, it may be most likely that both types of zinc
carboxylate contribute to the broad n
a
COO

band around 1570–
1590 cm
1
that we nd in ZnO-LO and in historical zinc white
paints. It is oen observed that zinc soaps containing mostly
saturated fatty acids form spontaneously in oil paint lms
containing ZnO, and since our system is so similar to paint
formulations, there is no reason why they should not be form-
ing at least to some extent in ZnO-LO. In fact, the shoulder in
the FTIR spectrum of ZnO-LO at 1536 cm
1
and the weak set of
long spacing peaks in the XRD results suggest that there is a low
concentration of crystalline zinc soaps present already.
Furthermore, if (semi-)crystalline zinc soap material is
nucleating somewhere in the paint system and subsequently
increasing in size, there must be transport of Zn
2+
ions from the
ZnO pigment particles to the growing zinc soap aggregate. Since
isolated ions are unlikely to exist in a relatively apolar medium
like a polymerized oil lm, as Zn
2+
moves through the paint
system it needs to be accompanied by a suitable negative
countercharge. The most obvious candidates for these coun-
tercharges are the carboxylic groups that are part of the polymer
network. Diffusion of metal ions in ionomer systems has been
Fig. 8 (a) FTIR spectra of zinc sorbate (Zn(C6:2), bottom) and a cured
mixture of zinc sorbate and linseed oil (Zn-pol, top), showing a broad
n
a
COO

band with a maximum at 1585 cm
1
. (b) Photograph of a
completely transparent thin yellow film of Zn-pol. (c) XRD trace of Zn-
pol.
1606 |J. Anal. At. Spectrom.,2015,30,1600–1608 This journal is © The Royal Society of Chemistry 2015
JAAS Paper
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

studied both experimentally and through molecular simula-
tions,
31–34
and was shown to occur through an “ion hopping”
mechanism in which metal ions are transferred between ionic
groups on different polymer chains. Crucially, this diffusion
process only occurs at an appreciable rate when the polymer is
heated above its glass transition temperature, when the polymer
chains are mobile enough to bring together their ionic groups to
a distance where the transfer of ion from one carboxylic group
to the other is feasible.
We conrmed to capability of Zn
2+
ions to exchange between
carboxylate groups in linseed oil experimentally. Two days aer
mixing of two clear linseed oil solutions, one containing ZnUFA
and the other palmitic acid, crystalline zinc palmitate could be
isolated from the resulting turbid reaction mixture and identi-
ed with FTIR spectroscopy. This simple pilot experiment
shows that even in an oil environment, the zinc-carboxylate
bond may be broken and the precipitation of zinc palmitate
drives the exchange of Zn
2+
ions from dissolved unsaturated
fatty acids to saturated fatty acid chains.
If ionomers can indeed be considered valid model systems
for aged oil paint, the glass transition temperature of oil paint
and any factor that inuences this physical property (e.g. rela-
tive humidity, solvent absorption, pigmentation) could become
a crucial parameter for the rate of metal ion diffusion and
therefore metal soap growth. Further investigations into the
preparation of ionomeric systems closely resembling a poly-
merized oil network and the analysis of these systems to study
the relation between diffusion processes in ionomers and oil
paint lms are the topic of a forthcoming publication.
4 Conclusions
In samples of zinc white paint layers, different types of zinc
carboxylates are found. A species characterized by a sharp n
a
-
COO

band at 1536 cm
1
corresponds to crystalline zinc soaps
of saturated fatty acids, but the nature of a second type char-
acterized by a broad n
a
COO

band with a maximum between
1570 and 1590 cm
1
has never been experimentally demon-
strated. We have shown that neither variations in the compo-
sition of zinc soaps (e.g. mixtures of fatty acids of different chain
length or number of double bonds, or mixtures of metal ions)
nor fatty acids adsorbed on pigment particles can account for
the broad vibration band. FTIR and XRD analysis on model
paint lms, however, showed that the zinc carboxylate species
must be amorphous, and initial experiments suggest that zinc
ionomers prepared from linseed oil may be accurate models for
mature binding media.
The exact nature of the carboxylate species—disordered
metal soaps or metal carboxylate functionalities covalently
linked to the polymerized oil network, or both—is at present
unclear. However, it is important to make the distinction
between zinc soaps and the broader class of zinc carboxylates in
the discussion of chemical processes in paint lms. These
ndings represent a breakthrough in the interpretation of FTIR
spectra of oil paint samples. A broadening and shiof the n
a
-
COO

band of the metal carboxylate signies an oil paint system
where pigments have partially degraded but metal soaps have not
been able to crystallize (yet), i.e. an intermediate stage between an
intact and a strongly deteriorated paint lm. As such, FTIR
analysis provides important information on the internal condi-
tions of oil paint layers and the degree of degradation, aiding
conservation treatments of invaluable works of art.
Acknowledgements
The authors thank Ella Hendriks (van Gogh Museum) and
Muriel Gelddof (Cultural Heritage Agency, the Netherlands) for
making the sample of the van Gogh painting available. This
work is part of the PAinT project, supported by the Science4Arts
program of the Dutch Organization for Scientic Research
(NWO), and the leadART project, part of the Joint Program
Initiative for Joint Research Projects on Cultural Heritage (JPI-
JHEP).
References
1 M. Lazzari and O. Chiantore, Polym. Degrad. Stab., 1999, 65,
303–313.
2 C. Higgit, M. Spring and D. Saunders, National Gallery
Technical Bulletin, 2003, 24,75–95.
3 J. van der Weerd, M. Gelddof, L. Struik van der Loeff,
R. Heeren and J. Boon, Zeitschrif¨
ur Kunsttechnologie und
Konservierung, 2003, 17, 407–416.
4 M. J. Plater, B. De Silva, T. Gelbrich, M. B. Hursthouse,
C. L. Higgitt and D. R. Saunders, Polyhedron, 2003, 22,
3171–3179.
5 P. Noble, A. van Loon and J. Boon, ICOM Committee for
Conservation, 14th Triennial Conference Preprints, The
Hague, London, 12–16 September 2005, pp. 496–503.
6 M. Spring, C. Ricci, D. A. Peggie and S. G. Kazarian, Anal.
Bioanal. Chem., 2008, 392,37–45.
7 Z. Kaszowska, K. Malek, M. Pa´
nczyk and A. Mikołajska, Vib.
Spectrosc., 2013, 65,1–11.
8 K. Keune and J. Boon, Stud. Conserv., 2007, 52, 161–176.
9 K. Keune and G. Boev´
e-Jones, Issues in Contemporary Oil
Paint, 2014, pp. 283–294.
10 J. Sanyova, S. Cersoy, P. Richardin, O. Lapr´
evote, P. Walter
and A. Brunelle, Anal. Chem., 2011, 83, 753–760.
11 V. Otero, D. Sanches, C. Montagner, M. Vilarigues, L. Carlyle,
J. A. Lopes and M. J. Melo, J. Raman Spectrosc., 2014, 45,
1197–1206.
12 L. Robinet and M.-C. Corbeil, Stud. Conserv., 2003, 48,23–40.
13 J. J. Hermans, K. Keune, A. van Loon, M. Stols-Witlox,
R. W. Corkery and P. D. Iedema, ICOM-CC 17th Triennial
Conference Preprints, Melbourne, Paris, 15–19 September
2014, p. 1603.
14 G. Osmond, J. Boon, L. Puskar and J. Drennan, Appl.
Spectrosc., 2012, 66, 1136–1144.
15 C. Clementi, F. Rosi, A. Romani, R. Vivani, B. G. Brunetti and
C. Miliani, Appl. Spectrosc., 2012, 66, 1233–1241.
16 R. Mazzeo, S. Prati, M. Quaranta, E. Joseph, E. Kendix and
M. Galeotti, Anal. Bioanal. Chem., 2008, 392,65–76.
17 G. Gautier, A. Bezur, K. Muir, F. Casadio and I. Fiedler, Appl.
Spectrosc., 2009, 63, 597–603.
This journal is © The Royal Society of Chemistry 2015 J. Anal. At. Spectrom.,2015,30,1600–1608 | 1607
Paper JAAS
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online

18 J. J. Hermans, K. Keune, A. van Loon, R. W. Corkery and
P. D. Iedema, Polyhedron, 2014, 81, 335–340.
19 A. Lenz, L. Seleg˚
ard, F. S¨
oderlind, A. Larsson, P. O. Holtz,
K. Uvdal, L. Ojam¨
ae and P.-O. K¨
all, J. Phys. Chem. C, 2009,
113, 17332–17341.
20 I. Dreveni, O. Berkesi, J. A. Andor and J. Mink, Inorg. Chim.
Acta, 1996, 249,17–23.
21 F. Lacouture, J. Peultier, M. François and J. Steinmetz, Acta
Crystallogr., Sect. C: Cryst. Struct. Commun., 2000, 56, 556–
557.
22 van Gogh's Studio Practice, ed. M. Vellekoop, M. Gelddof, E.
Hendriks, L. Jansen and A. de Tagle, Yale University Press,
2013.
23 M. Soucek, T. Khattab and J. Wu, Prog. Org. Coat., 2012, 73,
435–454.
24 N. Lah, G. Rep, P. ˇ
Segedin, L. Goliˇ
c and I. Leban, Acta
Crystallogr., Sect. C: Cryst. Struct. Commun., 2000, 56, 642–
643.
25 P. Taheri, T. Hauffman, J. M. C. Mol, J. R. Flores, F. Hannour,
J. H. W. de Wit and H. Terryn, J. Phys. Chem. C, 2011, 115,
17054–17067.
26 T. Ishioka, K. Maeda, I. Watanabe, S. Kawauchi and
M. Harada, Spectrochim. Acta, Part A, 2000, 56, 1731–1737.
27 S. Kutsumizu, K. Tadano, Y. Matsuda, M. Goto, H. Tachino,
H. Hara, E. Hirasawa, H. Tagawa, Y. Muroga and S. Yano,
Macromolecules, 2000, 33, 9044–9053.
28 M. Coleman, J. Lee and P. Painter, Macromolecules, 1990, 23,
2339–2345.
29 B. P. Grady, Polym. Eng. Sci., 2008, 48, 1029–1051.
30 S. Kutsumizu, H. Tagawa, Y. Muroga and S. Yano,
Macromolecules, 2000, 33, 3818–3827.
31 A. M. Castagna, W. Wang, K. I. Winey and J. Runt,
Macromolecules, 2011, 44, 2791–2798.
32 G. J. Tudryn, M. V. O'Reilly, S. Dou, D. R. King, K. I. Winey,
J. Runt and R. H. Colby, Macromolecules, 2012, 45, 3962–
3973.
33 L. M. Hall, M. J. Stevens and A. L. Frischknecht,
Macromolecules, 2012, 45, 8097–8108.
34 K.-J. Lin and J. K. Maranas, Macromolecules, 2012, 45, 6230–
6240.
1608 |J. Anal. At. Spectrom.,2015,30,1600–1608 This journal is © The Royal Society of Chemistry 2015
JAAS Paper
Published on 11 May 2015. Downloaded by Universiteit van Amsterdam on 26/06/2015 15:43:09.
View Article Online